Abstract 839: Microsatellite instability error correction in cell-free DNA sequencing

Background: Microsatellite instability (MSI) is often associated with poor cancer prognosis and is considered a hallmark in certain types of cancer such as Lynch Syndrome. MSI status has been approved by the FDA as a pan-cancer biomarker to stratify patients who exhibit clinical response to immune checkpoint inhibitor therapy. Circulating tumor DNA (ctDNA) is a novel, non-invasive and real-time approach to study various types of cancers including those that have tissue availability limitations. However, overwhelming majority of circulating cell-free DNA (cfDNA) is from healthy tissues, with a low ctDNA fraction. Another complicating factor is the generally elevated sequencing error rate in repeat regions where microsatellites (MS) are observed compared to non-repeat regions. Hence, MSI status inferred from ctDNA tends to be more error prone. Thus, it is important to differentiate signal from noise by error profiling to obtain accurate MSI status. Methodology: Here we present patterns in error rate that we observed in repeat sites and a method that leverages these observations to detect MSI status in ctDNA samples. We used 202 healthy cfDNA samples generated via TruSightTM Oncology 500 ctDNA, a research use assay (Illumina, Inc. San Diego CA), and evaluated more than 112000 repeat sites in an exploratory analysis. UMI collapsed reads were used to generate the read distribution at specific sites; the error rate is the fraction of reads with non-reference repeat counts. Based on this analysis, we developed an algorithm that generates the repeat length distribution for each MS site using duplex reads and then uses an information theoretic approach to determine the stability status of a specific locus. Final MSI score for a particular tumor sample was defined as the sum of distance between repeat length distribution of MS sites in a cohort of normal samples and the tumor sample. To assess the performance of our method, we applied it to cfDNA samples with known MSI status and titrated MSI-H cell lines with varying tumor fractions to obtain the MSI status. Results: The exploratory analysis revealed that error rates differed by 10 fold based on the read support in duplex sequencing; error rates increased as the repeat unit size of an MS site increased; for MS sites with the same repeat length, the error rate declined with higher repeat shift; the error rate in dinucleotide repeat sites was around 10 fold higher than mono nucleotide sites. We achieved 100% overall agreement in MSI status between 136 matched FFPE and cfDNA samples. For titrated MSI-H samples with low tumor fraction, our method attained 100% sensitivity at 0.625% MSI-H content titration into a microsatellite stable (MSS) background. Conclusion: Our analysis demonstrates that utilizing read characteristics from sequencing data leads to better prediction of MSI status and using this information our algorithm accurately predicts MSI status in cfDNA samples with wide ranging tumor content. Citation Format: Divy S. Kangeyan, Shile Zhang, Sigrid Katz, Brian Crain, Janel Lee, Alex G. Mentzer, Charlene Echegaray, Jennifer Silhavy, Weixin Wu, Tingting Jiang, Chen Zhao, Sven Bilke. Microsatellite instability error correction in cell-free DNA sequencing [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 839.